ID: 2.2/09.B
Physically-based hydrological models provide high-quality simulations of the water resources components, but running them for new regions is computationally expensive and time-consuming. Transfer learning and upscaling strategies aim to extend the prediction task of a machine learning model to other regions, with or without the same climatological conditions. In other words, once a model has learned the underlying dynamics in one place, we want to know: can it be reused elsewhere, instead of starting from scratch? This webinar walks you through the results of the STARS4Water upscaling and transfer learning work, from local case studies to continental-scale applications, and what these results actually mean for water management.
Learning Objectives
- Discuss main results obtained from upscaling and transfer learning across several case studies
- Application of ConvLSTM models over different regions
- Understand the limitations of this approach, and where it adds the most value
Target Audience
- Decisions makers, modelers, scientists
Keywords
Machine learning model, model training, upscaling, water resources assessment
Related Resources
[Category: 2.2 / Level: 2]
Presenter:
Dr. Leandro Avila, Forschungszentrum Jülich (FZJ)
For further questions please contact us via this form https://stars4water.eu/contact/